SIDS does not occur evenly throughout the year. There are well-known patterns associated with certain seasons and times of day. There is also a suggestion in the epidemiological literature that there are spikes associated with particular dates on the calendar, including today.
Because the other two dates with spikes are New Year’s Eve and April 20 (cannabis celebration day), it may be that on dates on which some parents consume psychoactive substances heavily, those parents are less attentive to their infants, or perhaps even endanger them (e.g., rolling over on top of them during co-sleeping). But if this is the mechanism, why aren’t there spikes on Christmas and Thanksgiving, dates upon which many American consume substances in an excessive fashion?
A further mystery: Why July 5 and not the holiday itself, July 4th? Are hungover parents the problem rather than the acutely intoxicated? Are the deaths not discovered until the next day because the parents don’t notice until then, or, not recorded by understaffed hospitals until the next day?
And finally, before one engages in too much story telling and explaining, it has to be said that SIDS is a mercifully rare event and any study of it thus deals in small samples. Studies of small samples can fool us into thinking an effect is there when there isn’t (see here and here). Perhaps the three date spike of SIDS death is just chance, and we see meaning in it that doesn’t in fact exist.
SIDS is not unique to the United States, but celebrating July 4th is. I understand Bastille Day and Guy Fawkes Day are similar in France and the UK, respectively. Do SIDS deaths in France and the UK show no spikes on 7/4 but similar spikes on their Big Nationalism Days? If so, then there is another bit of evidence favoring something in caretaker behaviors. If they show displaced peaks entirely, it’s probably coincidence.
@Dennis: That is really smart. Canada would be a good comparison point as well.
In the SIDS article, the New Year spike was the only one classified by the authors as a statistical outlier; the spikes on 5 July and on marijuana day were not statistical outliers. The New Year spike was recorded on 26 of the 29 years in the survey data (i.e., 26 positive residuals); the number of positive residuals for the other two dates was not reported, but were not considered by the authors to be outliers.
There is a dramatic difference between the graphic displays for SIDS by calendar date and the graphics for day of the week. This illustrates the law of small samples that Keith refers to; New Year comes but once a year, but there are 52 Saturday nights a year, which means you can look at a simple bar graph and not at a locally weighted scatterplot soothing polynomial line.
I suspect that Keith is correct in thinking that the spike occurs on July 5 rather than July 4 because dead infants are discovered on the following morning. That is why January 1 and not December 31 is where the spike occurs in the calendar date data and also why the bar graphs show higher numbers on Saturday and Sunday; people go to bed sober on Sunday and the graph for Monday is lower than for the two weekend days.
What is thought-provoking indeed is that infant deaths from other causes appear not to have a weekend effect and actually decreases on the weekend. SIDS, but not other causes of infant death, appear to show weekend increases in frequency.
Ed Whitney wrote: What is thought-provoking indeed is that infant deaths from other causes appear not to have a weekend effect and actually decreases on the weekend. SIDS, but not other causes of infant death, appear to show weekend increases in frequency.
As we know that consumption of many substances is linked to impulsiveness and aggression, one explanation that might be offered is that some intoxicated, frustrated parents are murdering their infants and those homicides are being misclassified as SIDS deaths. However, Harold Pollack did an elegant study (that sadly attracted little notice, maybe because it was complicated) analyzing the shape of SIDS death curves, which seemed to indicate that misdaignosed infanticide is not a significant explainer of the SIDS mortality rate.
http://onlinelibrary.wiley.com/doi/10.1111/j.1365-3016.2006.00693.x/abstract
This may get into what the French call “enculer les mouches,” but the study (published in 2010) used data from 1973 to 2006 for the day of week of death. A study done today could probably get data through 2009 at least.
The importance of sleep position as a risk factor for SIDS was not known in 1973. Once it was realized that babies need to be put to sleep on their backs, the Back to Sleep campaign began in 1994; more recently, this was expanded to develop the Safe to Sleep campaign. Death rates from SIDS declined beginning in 1994 as the percentage of babies placed to sleep on their backs began to rise. http://www.nichd.nih.gov/sids/Pages/sids.aspx has a link to a graph showing how steep the decline has been between 1998 and 2006. In 1997 the percentage of babies placed to sleep supine passed 50%. SIDS deaths per 1000 live births stood at 1.4 in 1988 and stood at 0.55 in 2006.
Some light could be shed on the epidemiological mystery by testing this hypothesis: one would predict that the effect of alcohol on SIDS would be different after the Back to Sleep campaign began. When no one knew any better (in 1988 only 13% of babies were put to sleep on their backs; in 2006 the figure was 75.7%), the effect of parental inebriation would be predicted to be smaller than when parents are expected to know better. The “weekend” effect would be predicted to be smaller as well. After, say, 1997, the weekend effect should be greater than in previous years, because intoxication will be more strongly associated with mistakes in infant sleep safety.
The numbers ought to be large enough to fit a regression model for the weekend effect and test for an interaction by period. When, prior to 1997, even the most sober parents were putting babies to sleep in a prone position, factors associated with alcohol consumption, the risk difference for infants of sober vs. inebriated parents should be less than in subsequent years.
It makes sense to me that there wouldn’t be spikes on Christmas and Thanksgiving, because those are primarily family holidays. There is probably both less overindulgence in substances and more extended family around to compensate for a distracted/incapacitated parent. People don’t get drunk/high on Christmas, do they? Aren’t there little kids/parents around?
The other thing, is if the other spike is on New Year’s Day, that oddly correlates to July 5; New Year’s is similarly the day after an evening of overindulgence (New Year’s Eve).
Also, it couldn’t possibly have something to do with fireworks, could it?
@Hazel: That is a plausible moderator. We might even be more specific and say that when the grandparents are around, their kids constrain their drinking which benefits their kids’ kids.
As for substance use on those holidays, I am not sure if this is still true, but Thanksgiving for many years was the number one date on the calendar for wine consumption in the US.
The Thanksgiving wine statistic could be a result of more over-consumption by a significant individuals, but it could also relate to more (probably a LOT more) individuals who choose to have some wine that day?
“…by a significant number of individuals…”
Ken Rhodes: That is certainly possible. Here is some anecdata from the restaurant industry
http://www.nbcnews.com/id/10154949/#.UdhfNPmTjE7
I wonder how much of the “spike” is a result of Ed Whitney’s suggestion about when infants are found. If you assume that infants are put to bed early on these days, and not checked till late morning, you’ve effectively got an extra quarter-day or more of actual time in the reporting day. Meanwhile, the family holidays are ones where infants are going to be spending more time with parents and other caregivers, rather than less.
Be careful about data spikes on January 1. They can be caused by software or data entry clerks defaulting to “1/1″ for events where date within the year is unknown.
Every study I’ve seen points to breathing problems as the source of SIDS- whether it’s from belly-sleeping, poor room ventilation, or brainstem malformations (in the breathing control areas). So I would posit smoke from marijuana and barbecue as the sources.
Maybe fireworks (home or community) too?
Fascinating question, fascinating responses from so many smart people. Enjoyed this post so much.
We’re not likely to draw accurate conclusions from analyzing statistical spikes in a single aspect of some phenomenon. An infant can die from SIDSwhile lying in his father’s lap. Cringely has the right idea — more data is needed.
I find these to be rather interesting theories. The primary focus is that parents should be aware of the dangers of SIDS and know that there are ways to prevent or in the least help minimize their child’s chances. http://www.itsbaby.com/reducing-your-babys-chances-of-sids-risk/
An education campaign launched in 1992 to have healthy babies sleep on their backs is credited with a 50% decrease in the infant mortality rate in the U.S. from Sudden Infant Death Syndrome. But along with the decline has come greater awareness of a condition called positional plagiocephaly, in which an infant’s head is flattened or misshapen, from too much time in the back position in the first months of life.
[...]
47% of babies studied have flat spots on heads: Placing healthy babies on their backs to sleep reduces the risk of SIDS and takes precedent over concerns about positional plagiocephaly, a flattened head condition that generally can be reversed.